New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers* received 16 January 1977 V P Kolonits, T Strausz and M K...

481KB Sizes 0 Downloads 25 Views

New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers* received 16 January 1977 V P Kolonits, T Strausz and M Koltai, Industrial Research Institute for Electronics, Budapest, Hungary

We have examined the thermal oxidation of tantalum (nitride) layers prepared for resistor application in order to find a kinetic model suitable to describe both short- and long-term changes occurring in the range between 200 and 360C°. The oxidation process was followed by resistivity measurements and the results so obtained were checked against a specially devised coulombmetry and SIMS method.

Introduction

A number of basic papers ~t-Ti deal with the thermal oxidation of bulk tantalum metal. These supply data concerning the further diffusion of oxygen bound either at the surface or inside the tantalum. These experiments are based on weight increase measurements or on embrittlement measurements during oxygen uptake. These methods, due to their character, are appropriate for detecting oxygen uptake mainly at the first rapid stage of the process. In order to describe this first stage on the basis of these measurements, some authors have found linear-', logarithmic 1'5, cubic 6, or p a r a b o li c relationships between the thickness of the oxide layer formed and the oxidation time at a given temperature. At the time of the appearance of vacuum sputtered tantalum and tantalum nitride layers, their behaviour tinder heattreatment was repeatedly examined. In this field, H Basseches a'9 was among the first: he sought correlation between increase in weight and resistivity change occurring during oxidation. It was he, and later A Schrauer t°'~1, who published data on the resistivity changes occurring during the first stage of oxidation, and later Schrauer and others ta gave data on the long-term changes at a lower temperature (typically in the range of I00-175°C). To avoid the fact that the character of the first stage and very long-term heat-treatment cannot be described by the same relationship, Grove ~3, in the case of silicon, applied the following mixed parabolic equation: N 0 2 "[-

A x o = B(t + r),

(1)

where not being able to insert the first stage into this relationship (not even by including an integration constant r) allows for the fact, that, for some reason (and here he refers to the theory of Cabrera and Mott re, which essentially can be deduced from the accelerating effect of the diffusion potential), the beginning of the process is far faster than is allowed by the chemical reaction speed constant.

"'A" type layer: was deposited by a triode sputtering system at a partial N 2 pressure of 2.5 .~ 10 -5 torr, with a total pressure of I :.: 10 - 3 torr. The deposition was effected after 20 min ionbombardment and 20 rain pre-sputtering cycle on to the surface of strongly heated substrates of 7059 type Corning glass. The layers consisted of 380-580-980 t~ tantalum nitride. After sputtering, the cool-off of the layers lasted 1 h in argon gas of 99.99~o purity at a pressure of I00 mtorr. " B " type layer: was deposited in pure argon gas in the same apparatus and under similar conditions to "'A". "C'" type layer: was deposited by diode sputtering (dc) at a sputtering voltage of 5 kV on to the surface of unheated substrates. In all three cases, the contact layer was TiNiAu, deposited in a separate apparatus, from which the LID = 30 size ratio resistors were prepared by selective photolithography. During the cleaning which followed the photolithographical process, the resistor networks were dried at 120°C for 30 min. The resistivity of the resistors was measured with a six-character digital multimeter. The heat-treatments were carried out in air at atmospheric pressure. The number of resistors which were annealed and measured jointly was at minimum 12, typically 24 and an average was calculated. The specific charge of the layers were determined with an electrolytic cell of the arrangement shown in Figure 1. The defined surface area needed for the determination was achieved by photoresist masking. The power consumption of the voltage measuring circuit was ensured to be very small so that an error less than 2.5 ~o was obtained at the critical stage of the determination--during the oxidation of the upper

ii I b

E xpe r i me nt a l

q o I

For comparison, the effects of heat-treatment on three differently prepared layers were examined:

To-Loyer substrote--- ~ .

* Paper presented at International Symposium on Vacuum and Thin Film Technology, Uppsala, Sweden, Aug. 31-Sept. 3, 1976. Vacuum/volume 27/number 12.

! Ii

I

I

~=1-D:~-r---~ I I -:[- l -etectrotyte II

!

=

Figure 1. Oxidizing cell for coulombmetric layer analysis.

Pergamon Press/Printed in Great Britain

635

V P Kolonits, T Strausz and M Kolta# New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

I00 ,~. In the course of tile following stages, tile measurement data were corrected by an average value. The arrangement was verified with the oxidation data obtained by oxidizing bulk, bright tantalum. A typical value of current density during the oxidation was 1.7 m A "~ cm -2. A Balmers SIMS was used. Following evacuation, the vacuum system was heated at 250°C for 7 h. Sputtering was with Ar + ions of 3 kV energy. The diameter of the primary beam was 2 mm. The surface was scanned at a primary beana current of I = 10 -7 A, while the depth composition determination was with I = 10 -6 A.

It can be readily seen from Figure 3 that the percentage o f resistance change is really inversely proportional to the initial thickness. Tile empirical data of Figure 3, when calculated applying equation (3), supplied at a given temperature for different film thicknesses identical d ~ values with a tolerance of !-200.~

~R/RH 0,3

do: 350~(200 £/n )

Discussion The R O = f ( d ) characteristic curve was plotted with samples prepared under similar circumstances but different sputtering times. The thickness of the layers was measured by the Tolansky method. These curves were needed in order to interpret the cross-sectional composition of the layers and to determine the depth of oxidation from the resistivity change caused by heattreatment. The empirical correlation can be seen in Figure 2, for " A " type layer preparation.

q

500

i/

-

ix

300

200

I00

X

-



X~

soo

doo

[q

Figure 2. Empirical correlation between layer thickness and resistivity for a given sputtering arrangement.

It can be seen from the curve that, in the course of further calculations, the first 200/~, can be treated as an insulator. The value of Ado, that is, the thickness change of the conductive layer due to heat-treatment can be calculated with the help of do--d~o which, in its turn, can be determined from R o ( ~ q o - I ) value of the unannealed layers with the help of Figure 2. The calculation is as follows:

Ro R.

d . - d= -do - lop

Ro -

-

RH

Rn

(2)

d n - doo - (do - d~o) ~

-Ad =

do - d ~

Ro d Ad 1 -- R'---n"~ do - d'-"'-'~= ---7' do

-

do - do~

(3)

(4)

where do c stands for the conductive portion of the initial do layer thickness.

636

o,1 ~

= • ~,

8

12

16

(lOOn/a)

do= 950 ~k(40n/o) 20

2~

28

hours

Figure 3. The effect of layer thickness on the resistance change ("A" type layer, heat-treatment at 250°C).

for 120 resistors. (The same good fit applies naturally to the " B " and " C " type layers too.) In order to verify the /Sd c values thus determined, we examined the thickness decrease of the metallic tantalum (or tantalum nitride) caused by equal c h a r g e - - f o r unit surface a r e a - - i n the case of untreated and differently annealed layers (that means the speed of the voltage increase at the anodic oxidation of tantalum layers). We succeeded in reproducing the slope of 0.33 V s - ' given in the literature for a current density of 1 m A c m -2 in the case of bulk tantalum sputtered in pure argon gas, which corresponds to the charge exchange process Ta -se ~ TaS+ and to the activation energy demand determined by Vermilyea '5, and made somewhat more accurate by others '6. F r o m the a b o v e data, 0.33 V × 10 -2 As cm - 2 = 0.33 V x 10 -a C o u l o m b cm -2. If a higher voltage increase is observed for the same charge quantity, it can signify either a smaller density or a higher oxidation state of the tantalum. Without counting the size change of the elementary cell, we can expect for the following processes: Ta2+

-3e > TaS+ = 0.55 V x 10 -3 C o u l o m b cm -2

Ta3+

-2e ~ TaS+ = 0.825 V x 10 -3 C o u l o m b cm -2

Ta4+

-e

~ T a s+ = 1.65V x 10 -3 C o u l o m b c m -2.

Apart from the different unit cell sizes corresponding to the different compositions, the different ionizing powers changing specifically had to be taken into account, so the above data only for reference. Experimental data can be found in the literature for tantalum nitride layers sputtered under different nitrogen partial pressures. This literature ' t , on the basis of X-ray structure investigations, correlates with the value of 0.52 V x 10 -3 C o u l o m b c m - 2 : the hexagonal Ta2N formula. F o r comparison purposes, in Figure 4.1 can be seen typical C o u l o m b metric curves for pure metallic tantalum and for layers sputtered according to methods " A " , " B " and " C " , respectively. Figure 4.2 shows the layer structures presumed. In the case of " C " type layers, where the films were sputtered on to unheated glass substrates, changes in the forming speed cannot

V P Kolonits, T Strausz and M Koltai: New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

b'Co~gcr6j

/



antCt u m " A

l • / bulk TQ

ayer

oxlde

!.-__ i il]i i.,_ :~, 3

~-,j X

% 5 -~.

Figure 4.1. Coulombmetric data of the pure metallic tantalum and of differently sputtered tantalum layers. 2. Cross-sectional picture of the examined layers corresponding to the A-B curves in Figure 4.1. be observed through the cross-section only in the thin airoxide layer. In the case of "'B" type layers, apart from the steep starting value caused by the air-oxide, the thick strongly oxidized or low-density zone near the glass side can also be found. In case "'A", the plateau shows a forming speed corresponding to Ta2N, the air-oxide and the substrate-side oxide exhibit values similar to those measured in case "B". Preparing

the SIMS spectrum of samples made by method "'A", the cross-sectional pattern shown in Figure 5 can be obtained. In the oxidized zone, the spectrum shows not only an increase of oxide with oxygen concentration, but tantalum and tantalum nitride increases as well; this, however, is the result of the higher ionization efficiency which increased usually by two orders of magnitude due to the increased oxygen present. The following figure shows that the upper oxide layer contains-apart from the four ions of Figure 5--further ions. Only those peaks were assigned on the figure which were characteristic for the oxidation degree of the layer and not for its impurity. Taking into account the time element of the process, Figure 6 should represent a state corresponding to a depth of 10 ,~. in a zone of -.'_-5/~. The explanation of the curve "'A" of Figures 4.1 and 4.2 is modified on the basis of Figures 5 and 6, in the sense that the zone closest to the substrate is probably of lowest density, that is it possesses an island structure. Thus, the increasing forming speed (Figure 4. I ) and the decrease of all intensities towards the substrate surface (Fig. 5) can be reconciled. By including these supplementary facts, the structural picture drawn by the coulombmetric curve for sputtered oxidized tantalum, tantalum nitride layers can be regarded as established. In the following, when evaluating the coulombmetric layer structure curves, it was considered that the voltage level of the oxidizing cell at the turning-on point corresponds to 16.5 ,~ oxide per V or to be more exact, corresponds to 16.5 ,~ per V of tantalum nitride of Ta s+ oxidation level, which is produced when oxidizing 5.7 A per V metallic tantalum. The following figures demonstrate how the forming speed of the layers prepared at different temperatures change starting from the air-tantalum (nitride) interface. It can be seen that heat-treatment changes the thickness of the barrier layer, which is followed by a tantalum or tantalum nitride layer of decreasing oxidation level. In Figure 7, those AU, + ~ U s points can be observed, which correspond to the ~ d c calculated from the resistance changes. Figure 8 shows that, at 360°C; this partially oxidized zone represents a large part of the whole thickness, and after 100 h practically no unoxidized

TaO"

Ta" cps

TaN"

T~"

SUBSTRATE

20

minutes

Figure 5. SIMS analysis of a tantalum nitride layer corresponding to the curve of Figure 4.1 (A). 637

V P Kolonits, T Strausz and M Koltai: New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers cps

16 1C

181

O-

To

* 197T00 +

28

N~-

T95

ri

2~3TaO*

m/e t.0

-

30

20



!0

20

tantalum nitride can be found. With this method for the layers of "A" and "B" type, the oxidation zone change along the substrate surface can be followed with less accuracy than on the air-side surface due to the ever-increasing parallel resistance. The layers sputtered by method "'C'" do not have intermediate zones on the substratc surface before heat-treatment. The thickness of the oxide step at some points was verified with the Talystep method: thus, the values of -~Ua for Figure 7 and ± U3 for Figure 9 were obtained well inside the accuracy limits for both methods (Talystep and forming speed measurements). The oxide thickness after 112 h annealing (Figure 8) could not be determined accurately due to the interdiffusion

Figure 6. SIMS spectrum of the upper (oxidized) layer of annealed tantalum nitride (I= 10- ~ A scanning current).

395~,

L,eooa41 [ . ~V__]t •

1,5

z%V

t,O •

/1:

ou~

c_~



250"C

~

aU~ ±~ aU 3

200"C

~00~

250 h

250 h

50

100

IV.

Figure 9. The cross-sectional coulombmetric characters of "B" layers before and after annealing.

Figure 7. The cross-sectional coulombmetric layer analysis characters of "A" layers after different annealing.

, spuUered

:, 4 hrs • 4hrs

o

/112h

1,o

/

Ta

250 °C 360°C

,I

h .

,,~3h

k

A~--

360'C

& .............,.~_ _

oh

Oh

tr

Figure 8. The effect of the 360°C annealing on the coulombmetric cross-sectional characters of"A" layers. 638

Er

'5(~ [ V]

Figure 10. The cross-sectional coulombmetric characters of "C" layers before and after annealing.

V P Kolonits, T Strausz and M Koltai." New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

of the tantalum nitride with the evaporated masking material which could not be removed by a well-defined etching. Calculating the changes of d - - w h i c h in the following arc symbolized by x -- wi t h the help of equation (4) from the measured resistance changes and plotting them as a function of time at different temperatures, the relationship in Figure 11 can be

t A d

360 °¢

:-

OoC

Figure 12. The empirical values of dx/dt =f(x) function at 40 #2 _~(A), 100 .(2 - - - t (O), and 360 #2 - - ~ (.:). Annealing at 2000°C. changes are also far bigger). This theory inspires another explanation, which essentially initiates from the diffusion theory of solids, where the diffusion is influenced by a potential field (chemical or electrochemical). Examining in detail on the basis of Evans ~7, those marginal cases, which by neglecting certain factors, lead to linear, logarithmic, parabolic or mixed-parabolic relationships, applying different presentations the speed data calculated from the measurements were observed in order to determine the most fitting mathematical formula and physical phenomenon. The general solution t7 for the oxide layer formation speed is:

300°C

dxo = A exp

d, 2~0°_C

+

-if-

- exp

.

(6)

XO t

:.-

...

,.

.

.

.

-

.

.

.

.

.

..

where A is ratio factor, U is chemical or electrochemical potential, Xo is thickness of the formed oxide layer, a is the size of a jump unit in the layer, W is activation energy necessary for the thermodiffusion of the oxidant. In a simplified form:

/boo,s/

Figure 11. The decreases of the conducting Ta-nitride layer thickness for different annealings ("A" layers). obtained. Although, at a first glance, the curve suggests that the processes can be described by Grove theory, further calculations show that no such A and B constants can be found which are valid for both the beginning and the slowing-down stage of the process. This difficulty can be overcome by supposing two parallel oxidizing reactions which both decrease the thickness of the metallic zone, and that the A and B constants of these reactions are very different. In order to follow this better, instead of the curve x=f(t) its differential form

0 -

dx

A

The simplified forms of equation (7):

dt

2x + B

dx---2 = K A (e ° - e -°) dt

(7)

where K = exp

(8)

Uaq x kT

(9)

(5) d.vo

should be observed. Figure 12 shows the speed curve corresponding to 200°C for " A " type layers. It can be seen that there are no A and B constants which give such a curve. It could be supposed, however, that, for example, the air-side and glass-side oxidation occur jointly, where different A and B constants regulate the two processes. Unequivocal proof of such further substrate-side oxidation could not be found for " A " and " B " type layers (while, in the case of " C " type layers, the forming speed underwent notable change so the resistance

-

-

dt

=

K A exp

(0)

(10)

give the so called inverse-logarithmic cLx'° = 2 K A 0 dt

(1 I)

the parabolic formula by the series development of exp 0 -exp (--0) on neglecting 0 ~ and the higher level members. In order to identify equations (10 and (11), in Figures 13-18 639

V P Kolonits, T Strausz and M Koltai: New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

. /

/

\

,/ .;r

\

i,L] ",:5

3,

'

-.

1/~

2,

tg

Figures 13 and 14. The dependence of the rate of conducting layer thickness decreases, calculated from R O changes, on parameters I/x and log x, respectively, at 200°C annealing (layers type "A").

i'g[ i CLC

- C

/

/

_!

/

- 2,C

':'~

l/x

1,

2,0

' "~" I g x

Figures 15 and 16. The dependence of the rate of conducting layer thickness decreases, calculated from R[:3 changes, on parameters l/x and Iogx, respectively, at 240°C annealing (layers of type "A").

3,0,~

0/3!

I/x

~,0



10 [gx

Figures 17 and 18. The dependence of the rate of conducting layer thickness decreases, calculated from RQ changes, on parameters l/x and log x, respectively, at 360°C annealing (layers of type "A"). 640

V P Kolonits, T Strausz and M Koltai: New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

can be seen the data corresponding to the resistance change due to heat-treatment for " A " type layers plotted logarithmically. In the calculations, instead of the increasing oxide thickness Xo, the proportional decrease in metal layer thickness .v was used. The conclusion can be drawn that equation (10) describes appropriately the middle stage of the process, while accepting the initial oxide layer thickness given by the forming speed measurements, it also explains a slower beginning for the process than that expected from an exponential correlation. The log dx/dt=f(Iog x) curves (Figures 13, 15 and 17) show that above a certain oxide thickness, the process slows down as compared to the exponential and assumes a ( - l ) directivity index, which corresponds to the approximate equation (l 1). In order to eliminate this break, the original ,4K [exp 0 - e x p ( - 0)] equation should be accepted, and a better approximation

would have to be applied than that of equation (1 I). Figure 19 shows the values obtained by including the cubic member and as can be seen, apart from the first, air-oxide determined stage, which was verified numerically, the approximation sufficed to eliminate the breakdown of the log x representation. This was considered to signify that equation (7) is the more accurate description of the physical phenomenon occurring during the process of thermal oxidation. We examined how the data described in the literature for tantalum and tantalum nitride fit into this theory. The literature data 1° plotted by our relationship can be seen in Figures 20-21 while those of literature I~ are shown in Figures 22-23. It can be seen that these data give similar characteristics to those of our own measurements. These results are in good agreement with the fact that the long-term examination of the layers gives data accommodating the parabolic equation m2.

:oral ~[dtj

l

0~-

\ -~o i I 0.0

0,I

I/%

-Z0

QO

I,C

2n

:3~) Ig xo

Figures 19 and 20. The oxidation data of tantalum (relating to 240°C) taken from ref. 9 and displayed in the system of Figures 13-18.

2~ I

'gI 7

rdxl g[~j

J ~ 3,C

C,0-

- \C

½ ,.U 0,O

\ 1~:,

/ -I(

o /

oo

Igx ~, ~0 1.0 T

Figures 21 and 22. The conducting film decrease data calculated from data given in ref. 10. The data are displayed in the system of Figures 13-18.

0

ID

Figure 23. The plot of log ((Ix/d/) against log ( I / x 4- Ka/xa), i.e. taking into account the 4th term of the expansion: Ka/x a. 841

V P Kolonits, T Strausz and M Koltai: New reaction kinetic aspects of the thermal oxidation of sputtered tantalum (nitride) layers

Conclusion

References

T h e oxidation processes occurring d u r i n g the a i r - a n n e a l i n g of sputtered t a n t a l u m nitride layers were examined. The process was followed by resistance measurements, which m e t h o d is a p p r o p r i a t e to show very small changes, while m a k i n g a great n u m b e r of m e a s u r e m e n t s a n d averaging them. T h e results were verified by m e a n s of c o u l o m b m e t r i c , S I M S a n d oxide step height measurements. T h e results o f these m e a s u r e m e n t s could be best interpreted by observing the relationship between the speed d a t a a n d the oxidized layer, a n d t h r o u g h these observations a u n i f o r m m e c h a n i s m could be supposed when accepting the validity of solid-state diffusion in a (chemical) potential field.

D A Vermilyca, ,4eta Met 6, 1958, 166. : R C Pelerscn, W M Fassel and M E Wadswordl, J :WetaLv (Tran.~ ,41ME) 200, 1038. 3 A Gulbarsen, Rer Sci Instrum 15. 1944, 201. '~ H D Gebhardt, H D Seghezzi and A. Stegherr, Z Metallkunde 48, 1957, 624. -~J. T. Waber, G E Sturdy, E M Wise and C R Topton, J. Tratts. Electrochem Soc 99, 1949, 12 I. 6 j T Waber, J Chem Ph),s 20, 1952, 734. 7 E A Gulbransen and K F Andrew, J Metals Trans A I M E 188, 1950, 568. s H Basseches; IRE Trans Compo/tent Parts, June 1961.5 I. '~ H Basseches, J Elec/rochenl Soc 109, 1962, 475. o A Schrauer. Siemens-Bauteile-htformationen 9, 1971 H ft, 9. J~ A Schrauer, Vortrag zum 4. Mikroelektronik-Kongress in M/.inchen vom 9-1 I, November 1970. ~-" F Huber and D Jafee, Proc 20th Electr Comp Conf(1970). t 3 B E Deal and A S Grove, J Appl Phys 36, 1965, 3770. =., N Cabrera and N F Mott, RepPro.qrPhys 12, 1948, 163. z.~ D A Vermilyea, Acta h4et I, 1953, 282. ~o L Young, Proc Roy Soc A, 244, 1958, 41. ~7 U R Evans, The Corrosion and Oxidathm o f Metals. Edward Arnold, London (1960)

Acknowledgement

T h e a u t h o r s are grateful for the assistance of Professor D r Jfinos G i b e r a n d others, for p e r f o r m i n g the S I M S measurem e n t s a n d D r G y 6 r g y Bajor for p e r f o r m i n g the Talystep measurements.

642